Comparison of Self-Localization Methods for Soccer Robots

This paper presents a comparison of two localization algorithms for tiny autonomous robots in a well known but highly dynamic environment. Position knowledge is gained through collecting information about distinctive features in the environment with a stereo vision system and comparing it to a model of the world. The model consists of a map of the static environment and information about moving objects. Based upon this model, the sensor data is used to generate a hypothesis of the position of the robot in the real world. A robot soccer game played by small, autonomous robots is the test-bed for this work. Constraints such as small size of the robot and the dynamic nature of the environment has to be taken into account while developing any solution for the localization. Simulation results show that for the given application, the extended Kalman filter based method is comparable in performance to the particle filter based method, although the particle filter has high time complexity.